Mistral’s New ‘Le Chonk’ AI Model Is Big, Open and Built for Agents
Mistral Large 4 is a state-of-the-art model for cyberdefense, manufacturing, finance-related tasks and multimodal use.
Mistral AI has unveiled its newest AI model, Mistral Large 4, which is both powerful and open-source. This model, affectionately dubbed "le Chonk," is designed to excel in text and image processing tasks. The company claims that it outperforms both European and US models, while also rivaling Chinese competitors in the open-weight sector.
Mistral Large 4 has been specifically engineered to tackle cybersecurity challenges, including log inspection, code auditing, and the identification of new vulnerabilities. It also boasts capabilities for offensive cybersecurity, aimed at identifying and addressing weaknesses.
Beyond cybersecurity, the model is engineered for agentic tasks such as planning, research, and complex enterprise management. In benchmark testing, it has demonstrated superior performance in areas like coding, analyzing, and processing pull requests. In addition, Mistral Large 4 is capable of advanced analysis of aerial and satellite imagery, identifying visual patterns and objects on maps, such as forest fires.
It also sets a new standard for grounding capabilities, anchoring its outputs in external resources to minimize inaccuracies and keep the information updated.
The model is multilingual, trained on over 160 languages, including EU official languages, Latin, and non-Latin scripts, allowing it to understand and generate responses in various languages. The model is set to launch on October 27, but a public preview is currently available for developers, cybersecurity experts, and state officials to evaluate and provide feedback. After this period, Mistral will conduct further reinforcement learning and fine-tuning to optimize the model's performance.
Written by urgent.news from CNET's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
